Crack Monitoring in Resonance Fatigue Testing of Welded Specimens Using Digital Image Correlation

Fatigue properties of welded details are commonly determined on small scale specimens, which can be efficiently be tested. The test consists in applying a cyclic load. Eventually a microscopic crack will initiate.

The crack will then grow and propagate through the specimen. The test is run until the specimen fades. The result of the test is the number of load cycles until failure for the pipe load level.

This find of failure is usually quite obvious, but how can the crack initiation be determined? An experimental approach using digital image correlation is presented in the following. The specimens used in the following test contain a multilayer between a 10mm and a 25mm plate.

The specimens are made of structural steel at 355. At the fatigue loading cracks are expected to form at the weld hole on the 10mm plate. For the digital image correlation a speckle pattern is applied on the specimen surface.

After the weld and the surrounding area have been cleaned of any dirt or oil, the pattern is applied using spray paint. The speckles are obtained by alternating layers of white and black paint. The nozzle is held at some distance from the specimen so that the spray forms fine speckles and not a close layer of paint.

The speckles should be as fine as possible at a magnitude of 0.1mm. The tests are run on a 200 kilo newton resonance testing machine. In this setup the cameras for the digital image correlation are placed above the specimen.

It is crucial to properly set the focus and aperture of the camera lenses. To allow for short exposure times, sufficient illumination has to be provided. Four LED lights were positioned close to the specimen.

In order to reduce reflections, polarization filters were applied on the lights and cameras. The tests are run at a loading frequency of 34 Hz, resulting in a period of about 29 milliseconds for each load cycle. The exposure time of the cameras should be a small enough fraction of this load period.

For the used setup an exposure time of 0.8 milliseconds has proven suitable. The cameras are triggered by the force signal from the testing machine. To compensate for the delay between trigger signal and actual image acquisition it might be necessary to set the trigger somewhat before the peak of the load signal.

The first load cycle is applied statically. At the maximum load an image of the weld is taken. The actual fatigue test for cyclic loading is then started.

The images are shot at predefined intervals of load cycles triggered by the force signal from the testing machine without interrupting the test. The interval should be chosen in order to obtain about 100 to 200 images over the duration of the test. This should be enough to determine crack initiation with the necessary accuracy, while avoiding excessive data acquisition.

The generation of beach marks is optional. It is applied here to verify the crack length detected by digital image correlation. Periods are a reduced load range, and introduced periodically over the duration of the fatigue test.

The decreased crack propagation during these intervals becomes visible in terms of semielliptical marks on the crack surface after completing the test. After the test, digital image correlation is evaluated to calculate strains in the loading direction of the specimen. The exact procedure will depend on applied software.

To calculate the strains the image from the first static load cycle is used as a reference, where all strains calculated for the successive images will be relative. The range of the control plot is adapted to suppress possible noise and evidence crack formation. Then run through the images acquired over the duration of the test.

Eventually the strain at the weld hole will begin to increase, indicating another crack is forming. Technical or microscopic crack initiation, whilst assumed constraint exceeded 1%of a length of 2mm. In the as-welded condition this specimen contains tensile residual stresses in the middle of the specimen, and compressive ones at the edges.

The crack is therefore expected to initiate close to the center line of the specimen. At this point a microscopic crack has formed. To validate observed crack length, the results are compared to beach marks generated during the test.

The crack growth is visibly slowed down during the formation of the beach marks. This specimen was stress-relieved after welding. Crack initiation is therefore not influenced by residual stresses.

Several cracks formed at different locations along the weld. They grew as indicated by the beach marks and eventually unified. The presented procedure using digital image correlation allows the technical crack initiation and monitor and crack propagation during fatigue tests.

It is applicable on resonance testing machines with high load frequencies, without interrupting the running test. Adopted on welded specimens it allows to cover the whole width of the specimen to detect cracks initiating at the weld hole.